[edit] some answers :p =

well, finaly i read this page, hum :)

i also noticed some emerging errors while using tmpfs, no idea what they are about.. no time to look it up. same on causing longer emerging-times, its mosty while using swap.. i guess

my testingmethod is really bad, using time insteam of genlop would be nicer. feel free to submit testresults :) greetz, inbreed

[edit] some test results

So, it looks like running it in RAM does seam to help, but not xorg in under 8 minutes or anything. I wrote the following file, testrun to run the different compiles for me, and umm ignore my bash, I'm still learning and didn't feel like putting a lot of enery into it.

#!/bin/bash
TIME=`which time`
FILE=results_${1}
 
emerge -f ${1}
echo "compiling tests with ${1}" > $FILE;
echo "Time results for \"emerge ${*}\"" >> $FILE &&
$TIME -a -o $FILE -p -- emerge ${*} ;
echo "Time results for \"./temerge1 ${*}\"" >> $FILE &&
$TIME -a -o $FILE -p -- ./temerge1 ${*} ;
echo "Time results for \"FEATURES=\"-ccache\"./temerge1 ${*}\"" >> $FILE &&
FEATURES="-ccache" $TIME -a -o $FILE -p -- ./temerge1 ${*} ;
echo "Time results for \"./temerge2 ${*}\"" >> $FILE &&
$TIME -a -o $FILE -p -- ./temerge2 ${*} ;
echo "Time results for \"FEATURES=\"-ccache\"./temerge2 ${*}\"" >> $FILE &&
FEATURES="-ccache" $TIME -a -o $FILE -p -- ./temerge2 ${*}
 
 
more $FILE

Where temerge1 is the file original file provided and temerge2 just mounts the tmpfs with the extra option: nr_inodes=1M I ran the tests in command line with no X running. oh my system has 1.5GB RAM. I just wonder, would having more compiler optimizations reduce the differance between emerge and temerge? I assume with more compilter optimizations a greater % of the time is spent in the cpu cycle rather than accessing hdd/ram. Anyways, here are the results.

compiling tests with mozilla-firefox
Time results for "emerge mozilla-firefox"
real 686.55
user 487.02
sys 404.97
Time results for "./temerge1 mozilla-firefox"
Command exited with non-zero status 1
real 550.46
user 488.11
sys 398.39
Time results for "FEATURES="-ccache"./temerge1 mozilla-firefox"
Command exited with non-zero status 1
real 1655.29
user 2418.64
sys 449.29
Time results for "./temerge2 mozilla-firefox"
Command exited with non-zero status 1
real 552.51
user 488.06
sys 400.66
Time results for "FEATURES="-ccache"./temerge2 mozilla-firefox"
Command exited with non-zero status 1
real 1631.52
user 2398.63
sys 451.98

--Aperion

[edit] script mod

I find it's useful to keep the tmpfs mounted in case of a emerge failure so I don't lose the whole source tree when the package is 90% compiled. Especially useful on big ebuilds.

#!/bin/bash
MEMSIZE=850M
mounted=false
 
. /etc/init.d/functions.sh
 
mounttmpfs() {
     mount -t tmpfs tmpfs -o size=$MEMSIZE /var/tmp/portage
     mounted="true"
}

compile() {
     einfo "emerging ${*}"
     emerge ${*}
     ret=$?
}

unmount() {
     ebegin "unmounting tmpfs"
     umount -f /var/tmp/portage
     eend $?
}

ebegin "Mounting $MEMSIZE of memory to /var/tmp/portage"
if [ -z "$(mount | grep /var/tmp/portage)" ]
then
     mounttmpfs
else
     eerror "tmpfs already mounted!"
     exit 0
fi
eend $?

compile ${*}
 
if [[ -n "$mounted" && ! $ret ]]
then
     unmount
fi

[edit] Other Benchmarks

Maybe for the testing, rather than just an end result, use genlop to show the results of an emerge and a temerge? For example a test of QT I just did.

 # genlop -t x11-libs/qt-4.1.2
 Without tmpfs: x11-libs/qt-4.1.2 
      merge time: 2 hours, 28 minutes and 37 seconds.
 With tmpfs: x11-libs/qt-4.1.2

Taking a small package and running it through time.

emerge 3.55.540  vs   2.55.249
       1.24.295  vs   1.24.272
       1.29.718  vs   1.29.810

[edit] TMPFS and Block Device merged as one device?

"I don't have a huge amount of memory, (512M) but the tmpfs method works great on the packages that do fit. What I was wondering is it possible to make a linear RAID with tmpfs and a hard drive, or perhaps modify unionfs to work with tmpfs and a drive directory in which it will switch to the drive directory when tmpfs is full. - Gary"

"If you create a swap file/partition, and have your system use that, then set the tmpfs to bigger than your ramsize, but smaller than your ramsize+swap, then when your ram is full because of tmpfs, the computer will use swap space as extra ram."

-Dantarion

Thanks, I figured that out a while later. Been working like a charm for a long time. I'm trying /tmp and /var/log today as well, especially for my new laptop.

-Gary

[edit] Updated location of functions.sh

My current version of portage (2.1.5_rc4) doesn't have /sbin/functions.sh, so the script won't run properly. The new equivalent of this file is /usr/lib/portage/bin/isolated-functions.sh.

- TeknoHog

Nope. According to this post the correct file should be

/etc/init.d/functions.sh

I've corrected the script.

-Thetamind

[edit] Top tip for us using flash drives.

Less writes = longer life.